«Automation and Informatization of the fuel and energy complex»

Simulation non-stationary hybrid model of a large scale gas transportation system with hardware acceleration of computations on graphic processing units

UDC: 519.673
DOI: -

Authors:

¹ NIIGazekonomika, Moscow, Russia

Keywords: non-stationary process, simulation, gas transportation system, neural network, differentiable programming

Annotation:

The growing importance of natural gas in ensuring energy security and reliability of energy supply to consumers brings to the forefront the tasks of simulating and planning non-stationary modes of the gas transportation system (GTS) operation of the Russian Unified Gas Supply System. With account for the high dimensionality of these problems and the computational complexity, their solution requires the use of rationally built mathematical models and algorithms that ensure a balance between the accuracy of the simulation results and speed. The author of the article proposes a new approach to the computer implementation of computational procedures for simulating non-stationary modes of operation of large GTS. The simulation algorithm is proposed to be based on a non-stationary model of gas flow in main gas pipelines developed by domestic scientists, which requires significantly lower computational costs compared to classical models in the form of partial differential equations. A distinctive feature of the present article is the use of a new paradigm – differentiable programming ("differentiable physics") – for software implementation of the calculation algorithm. The main characteristics of the developed auto-differentiable simulation calculation model, which is adapted for performing tensor parallel computing on graphics processing units (GPUs), are presented. It is shown that the proposed mathematical, software and hardware computation methods allow possible developing a new efficient computer technology for modeling (simulating) non-stationary operational modes of large GTS with hardware computations acceleration. An approach to creating hybrid models that combine "exact" physical models based on knowledge with machine learning models based on data is discussed. It is noted that differentiable programming is a link between classical modeling of systems and the use of machine learning models, in particular, artificial neural networks. The results of computational experiments are demonstrated, indicating the high performance of the developed algorithm when simulating the operational modes of real large GTS. Directions for further research on the creation of new algorithms for optimizing non-stationary operating modes of large GTS based on the proposed auto-differentiable simulation model are discussed.

Bibliography:

1. Modelirovanie transporta prirodnogo gaza v rezhime onlayn. Programmno-vychislitel'nyy kompleks "Volna" / M.G. Anuchin, M.G. Anuchin, A.A. Anfalov [i dr.] // Neft'. Gaz. Novatsii. – 2017. – № 5. – S. 27–35.
2. Sardanashvili S.A. Raschetnye metody i algoritmy (truboprovodnyy transport gaza). – M.: Izd-vo "Neft' i gaz" RGU nefti i gaza im. I.M. Gubkina, 2005. – 577 s.
3. Sukharev M.G., Samoylov R.V. Analiz i upravlenie statsionarnymi i nestatsionarnymi rezhimami transporta gaza. – M.: Izdat. tsentr RGU nefti i gaza (NIU) im. I.M. Gubkina, 2016. – 399 s.
4. Sukharev M.G., Samoylov R.V. Modeli s sosredotochennymi parametrami dlya nestatsionarnogo techeniya gaza v magistral'nykh gazoprovodakh // Nauch.-tekhn. sb. Vesti gazovoy nauki. – 2022. – № 2(51). – S. 4–15.
5. Sukharev M.G., Popov R.V. Novaya metodika modelirovaniya nestatsionarnykh techeniy gaza v sistemakh gazosnabzheniya // Izv. RAN. Energetika. – 2015. – № 2. – S. 150–159.
6. Belinskiy A.V. Differentsiruemaya fizika – osnova tsifrovykh dvoynikov v neftegazovom komplekse // Avtomatizatsiya i informatizatsiya TEK. – 2024. – № 12(617). – S. 38–50.
7. STO Gazprom 2-3.5-051-2006. Normy tekhnologicheskogo proektirovaniya magistral'nykh gazoprovodov. – Vved. 2006–07–03. – M.: IRTs Gazprom, 2006. – 196 s.
8. Energoeffektivnye rezhimy gazotransportnykh sistem i printsipy ikh obespecheniya / A.M. Karasevich, M.G. Sukharev, A.V. Belinskiy [i dr.] // Gazovaya prom-st'. – 2012. – № 1(672). – S. 30–34.
9. Korel'shteyn L.B., Pashenkova E.S. Opyt ispol'zovaniya metodov global'nogo gradienta i dekompozitsii pri raschete ustanovivshegosya neizotermicheskogo techeniya zhidkostey i gazov v truboprovodakh // Truboprovodnye sistemy energetiki. Mat. modelirovanie i optimizatsiya. – Novosibirsk: Nauka, 2010. – S. 103–114.
10. Razrabotka i aprobatsiya metodicheskikh podkhodov i tsifrovykh tekhnologiy neyrosetevogo proksi-modelirovaniya ustanovivshegosya dvukhfaznogo techeniya mnogokomponentnoy smesi v sistemakh sbora i promyslovoy podgotovki gaza (na primere Chayandinskogo NGKM) / A.V. Belinskiy, V.A. Marishkin, V.V. Samsonova, P.V. Pyatibratov // Avtomatizatsiya i informatizatsiya TEK. – 2024. – № 4(609). – S. 44–59.
11. O novom metode tsifrovogo modelirovaniya nestatsionarnykh rezhimov techeniya gaza v magistral'nykh gazoprovodakh s primeneniem neyronnykh operatorov / A.V. Belinskiy, D.V. Gorlov, I.A. Pyatyshev, A.E. Titov // Gazovaya prom-st'. – 2024. – № 5(865). – S. 54–66.
12. O vozmozhnosti ispol'zovaniya tekhnologii mashinnogo obucheniya dlya modelirovaniya kompressornoy stantsii magistral'nogo gazoprovoda / A.V. Oleynikov, V.A. Shevchenko, A.V. Belinskiy, A.V. Maletin // Avtomatizatsiya i informatizatsiya TEK. – 2024. – № 6(611). – S. 35–45.
13. Korel'shteyn L.B. Sushchestvovanie, edinstvennost' i monotonnost' resheniya zadachi potokoraspredeleniya v gidravlicheskikh tsepyakh s zavisyashchimi ot davleniya zamykayushchimi sootnosheniyami // Mat. modeli i metody analiza i optimal'nogo sinteza razvivayushchikhsya truboprovodnykh i gidravlicheskikh sistem: tr. XVI Vseros. nauch. seminara, Irkutsk, 26 iyunya – 2 iyulya 2018 g. – Irkutsk: In-t sistem energetiki im. L.A. Melent'eva SO RAN, 2018. – S. 55–83.
14. Blondel M., Roulet V. The Elements of Differentiable Programming. – 2024. – 437 p. – DOI: 10.48550/arXiv.2403.14606
15. Runje D., Shankaranarayana S.M. Constrained Monotonic Neural Networks. – 2023. – 16 p. – DOI: 10.48550/arXiv.2205.11775
16. Ob obosnovannosti primeneniya, sovremennom sostoyanii i nekotorykh perspektivakh razvitiya neyrosetevykh modeley Edinoy sistemy gazosnabzheniya Rossii / N.A. Kislenko, A.V. Belinskiy, A.S. Kazak, O.I. Belinskaya // Avtomatizatsiya i informatizatsiya TEK. – 2022. – № 5(586). – S. 6–17. – DOI: 10.33285/2782-604X-2022-5(586)-6-17